cto-playbook
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFEREMOTE_CODE_EXECUTIONCOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [REMOTE_CODE_EXECUTION]: The skill instructs the agent to install the
uvtool by executing a shell script fromastral.sh, which is the official domain for a well-known technology provider. This is documented as a safe operation per trusted source rules. - [COMMAND_EXECUTION]: Directs the agent to utilize
uvxto run a security scanning utility (snyk-agent-scan) to audit other agent skills for vulnerabilities. This command downloads and runs code from a public registry as part of a security workflow. - [PROMPT_INJECTION]: Contains strong persona-defining instructions ("You are operating as a world-class CTO") and mandatory quality gates ("Non-Negotiable", "If in doubt, use this skill") to enforce engineering standards. These patterns are used for quality assurance rather than bypassing safety protocols.
- [DATA_EXPOSURE]: Lists specific local filesystem paths for agent skill directories (e.g.,
~/.claude/skills) to facilitate security scanning, which is a legitimate use of file path information in an administrative context.
Audit Metadata